Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated, it should be understood that and the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
As shown in Figure 1, the embodiment of the present invention provides a kind of lane sideline extraction method based on laser point cloud, it is described
Lane sideline extraction method based on laser point cloud the following steps are included:
S1, the three-dimensional laser point cloud that information of road surface is acquired by traverse measurement vehicle, and laser point cloud data is read out;
Points are less than threshold value points by S2, the set that the laser point cloud is divided into varying strength according to reflected intensity
Set is filtered, and carries out clustering to set, obtains active strength set;
S3, connectivity identification is carried out to the laser point cloud point in active strength set, finds the laser that line feature is presented
The linear connection subset of point cloud point;
S4, judge whether each linear connection subset is same line packetized elementary, and by the threadiness with same line packetized elementary
Connection subset is merged;
S5, the traverse measurement vehicle driving trace in conjunction with same position carry out lane line to fused linear connection subset
Identification.
The clustering is mainly clustered and is filtered to the point in cloud according to the reflected intensity of laser point, thus
It would be impossible to provide strength type to be selected as noise remove, and for the identification of subsequent lane line there are the point cloud of linear element,
Specific steps are as shown in Figure 2:
The step S2 include it is following step by step;
S21, the laser point cloud is divided by multiple and different intensity set according to reflected intensity;
S22, setting threshold value are counted, and the points of each intensity set and threshold value points are compared, if intensity set
Points less than threshold value count, then delete the intensity set, remaining strength set is combined into active strength set;
All active strength set are carried out clustering according to cluster intensity threshold by S23, setting cluster intensity threshold,
Until the reflected intensity of laser point cloud is respectively less than cluster intensity threshold in each active strength set;
S24, setting cluster threshold value points, by the active strength set after progress clustering, points are less than cluster threshold
The active strength set of value points is deleted.
Wherein, as shown in figure 3, the step S2 3 include it is following step by step;
S231, setting cluster intensity threshold calculate to be discriminated using all active strength set as a subclass to be discriminated
The average reflection intensity and reflected intensity mean square deviation of subclass.
If the reflected intensity mean square deviation of S232, subclass to be discriminated is greater than cluster intensity threshold, will be in subclass to be discriminated
Point according to reflected intensity distance center, and with the left and right one reflected intensity mean square deviation in interval of reflected intensity distance center
Subclass to be identified is divided into three subclasses by the rule of three point distances;
If the mean square deviation of the reflected intensity of S233, subclass to be discriminated is less than cluster intensity threshold, the subclass to be discriminated
Complete cluster;
S234, intensity threshold is clustered up to the reflected intensity of all subclasses to be discriminated is respectively less than.
Specifically, successively traversing all the points, and strong according to the reflection of point in the laser point cloud data obtained according to step S1
Angle value classifies to the point in cloud, is denoted as PCi, i=1,2 ..., n, wherein reflected intensity class PCiThe reflection of middle all the points
Intensity is identical.Remember reflected intensity class PCiIn points beAll points are Num in laser point cloud0。
Traverse reflected intensity class PCi, i=1,2 ..., n, if PCiIn pointsLess than threshold value points MIN_PT_
(MIN_PT_NUM is used to describe to constitute the noise reflection point number of road element expression to NUM, is a lesser numerical value threshold
Value, such as, but not limited to 100), then deletes the PCiIn all the points, and adjust point Yun to be processed and always count
Setting cluster intensity threshold CLUSTER__INV, using all active strength set as a subclass to be discriminated, meter
Calculate the average reflection intensity AVG_INV and reflected intensity mean square deviation STDEV_INV of subclass to be discriminated.
If the mean square deviation of the reflected intensity of subclass to be discriminated, which is greater than, clusters intensity threshold, i.e. STDEV_INV >
CLUSTER__INV.Then by the point in subclass to be discriminated according to reflected intensity distance AVG_INV-STDEV_INV, AVG_INV,
AVG_INV+STDEV_INV nearest principle, by subclass to be identified be divided into left (AVG_INV-STDEV_INV), in (AVG_
INV), right (AVG_INV+STDEV_INV) three subclasses.
Otherwise, if the mean square deviation of the reflected intensity of subclass to be discriminated, which is not more than, clusters intensity threshold, i.e. STDEV_INV≤
CLUSTER__INV, then the subclass to be discriminated completes cluster,
Wherein CLUSTER__INV is the threshold value of end of clustering, indicates that the reflected intensity of all the points in a subclass all compares
It is close, generally desired value should be selected according to the range of laser reflection intensity, so that whole clusters number is at more than ten or so, example
Such as, but not limited to, the point cloud for reflected intensity 60,000 or so, CLUSTER__INV=1000.
According to the method described above, until the reflected intensity of all subclasses to be discriminated, which is respectively less than, clusters intensity threshold, obtained institute
Having subclass is CCi, i=1,2 ..., m, subclasses C CiThe sum at midpoint is NumCi, total points of all subclasses are Num0。
Preliminary screening is carried out to each subclass that cluster obtains, on the contrary it will not be possible to be the subclass removal of Road, i.e., will own
Meet NumCi> α * Num0Subclass removal, wherein α be the coefficient for differentiating the expression of non-linear shape element, and value range is 0 to 1
Real number, but it is general should not be too small, such as, but not limited to α=0.33.
Wherein, as shown in figure 4, the step S3 include it is following step by step;
S31, connectivity identification is carried out to the point in active strength set, establishes the connection subset of each point.
S32, it carries out curve fitting to the connection subset of each laser point cloud point, identifies whether to be connected to subset for threadiness.
As shown in figure 5, the step S31 include it is following step by step;
S311, a point in active strength set is randomly selected, a connection subset is established based on the point;
S312, setting connection threshold value calculate in the active strength set all distances with the point less than described and are connected to threshold value
Point, and be added into the connection subset;
S313, pass sequentially through connection threshold value establish multiple connection subsets, until the active strength set in all the points belong to
In one of connection subset.
As shown in fig. 6, the step S32 include it is following step by step;
S321, it carries out curve fitting to the point for belonging to a connection subset;
S322, average distance of all the points in the connection subset with respect to matched curve is calculated;
If S323, average distance are less than connection threshold value, which is that a linear connection subset otherwise should
Connection subset is not linear connection subset.
Specifically, randomly selecting a untreated point in active strength set, it is denoted as P0, and be P0Establish new connection
Subset CON_0.
Setting connection threshold value, calculates all and P in the active strength set0Point distance is less than connection threshold value CONN_
Point { the P of THRESHOLDjJ=1,2 ..., k }, then by { PjJ=1,2 ..., k in it is all not connection subset CON_0 in points
Connection subset CON_0 is added.
It carries out curve fitting, and is calculated in the connection subset to the point in each connection subset, belonging to a connection subset
All the points with respect to the average distance of matched curve, be denoted as AVG_ERROR, if AVG_ERROR < CONN_THRESHOLD,
The connection subset is a linear connection subset, and otherwise, which is not linear connection subset.
Wherein, as shown in fig. 7, the step S4 include it is following step by step;
S41, a benchmark threadiness connection subset in active strength set is randomly selected, finds and matches in other connection subsets
Subset is connected to threadiness;
Point in S42, the linear connection subset of the pairing is connected to the average departure of the analytical expression of subset to benchmark threadiness
The threadiness of the same shape is expressed even from linear one for being connected to subset on the basis of less than threshold value is connected to, then matching threadiness connection subset
Logical subset.
S43, it all pairing threadiness connection subsets be connected to subset with benchmark threadiness merges, until all threadiness companies
Logical subset all complete by processing.
Specifically, randomly selecting a untreated threadiness in active strength set is connected to subset CON_I as reference line
Shape is connected to subset, is found in the linear connection subset CON_J:CON_J of all pairings for meeting following condition in other subclasses
The average distance that point arrives the analytical expression of CON_I is less than CONN_THRESHOLD, and CON_J is that an expression of CON_I is same
All threadiness for expressing the same shape with CON_I are connected to subset and merged by the linear connection subset of shape.
Wherein, as shown in figure 8, the step S5 include it is following step by step;
S51, a fused linear connection subset is chosen, loads the traverse measurement vehicle driving trace of same position;
S52, projection of the threadiness connection subset in traveling trajectory line is calculated, and calculates the length of view field;
S53, setting Lane detection threshold value, if the length of view field is connected to subset matched curve itself with the threadiness
The ratio of length is greater than Lane detection threshold value, then judges the linear connection subset for lane line;
S54, it carries out curve fitting again to the linear connection subset for being identified as lane line, obtains the shape line of the lane line
And width.
Specifically, linear connection subset to be identified after choosing a fusion, loads the traverse measurement garage of same position
Sail track.Then projection of the threadiness connection subset in traveling trajectory line is calculated, and calculates the length of view field, is denoted as
PRJ_LEN.The length of threadiness connection subset matched curve itself is denoted as FIT_LEN.If PRJ_LEN/FIT_LEN >
REC_THRESHOD is then identified as lane line.
Wherein, the REC_THRESHOD is Lane detection threshold value, and expression is the one of matched curve and driving trace
Cause property, generally close to 1 real number, such as, but not limited to REC_THRESHOD=0.85.
It carries out curve fitting again to the linear connection subset for being identified as lane line, the shape line of the lane line can be obtained
And width.
The lane sideline automatic extracting system based on laser point cloud that the present invention also provides a kind of, as shown in figure 9, the base
In the lane sideline automatic extracting system of laser point cloud include following functions module:
Point cloud data read module 10, the three-dimensional laser point cloud for acquiring information of road surface by traverse measurement vehicle, and it is right
Laser point cloud data is read out;
Efficient set obtains module 20, the set for the laser point cloud to be divided into varying strength according to reflected intensity,
The set that points are less than threshold value points is filtered, and clustering is carried out to set, obtains active strength set;
Linear subset obtains module 30, for carrying out connectivity identification to the laser point cloud point in active strength set, looks for
To the linear connection subset for the laser point cloud point that line feature is presented;
Linear subset Fusion Module 40, for judging whether each linear connection subset is same line packetized elementary, and will tool
There is the linear connection subset of same line packetized elementary to be merged;
Lane detection module 50, the traverse measurement vehicle driving trace for combining same position, to fused threadiness
It is connected to subset and carries out Lane detection.
Lane sideline extraction method and its system of the present invention based on laser point cloud, by being surveyed based on mobile
The laser point cloud for measuring vehicle acquisition carries out in high-precision electronic navigation map data element production process, according to the anti-of each laser point
Intensity is penetrated, unmanned very important lane line data element is automatically extracted with certain accuracy, is subsequent
The production of lane grade high-precision map provides basic lane shape data, to improve the acquisition of lane side line data element and life
The efficiency and accuracy of production, while being also greatly improved the efficiency of lane grade high-precision map producing.
Related terms of the present invention are explained:
1. high-precision electronic navigation map
Relatively traditional using road as the precision of basic element is the navigation map of meter level, the precision provided be decimetre even
Centimeter Level, using lane as basic element, towards unmanned and active safety application function next-generation digital navigation map.
2. traverse measurement vehicle
Laser scanner, panorama camera, high accuracy positioning equipment and High Accuracy Inertial equipment are installed, are capable of providing high-precision
Spend the measurement vehicle of location information.
3. laser point cloud
The laser point cloud that the laser scanner scans with location information and reflected intensity of traverse measurement vehicle acquisition obtain.
Point cloud is referred to as in text.As shown in Figure 2.
4. laser point cloud point
Each point in laser point cloud, referred to herein as laser point cloud point, also referred to as point.
5. lane sideline
Lane both sides printing sideline, text in also referred to as lane line, as shown in Figure 3.
6. reflected intensity set
The set of all the points with identical reflected intensity, referred to as a reflected intensity set in point cloud.
7. being connected to subset
Only in a specific subset (such as reflected intensity set) for cloud or point cloud, interconnected all the points are constituted
A subset.
8. threadiness connection subset
If the global shape of a connection subset shows line feature, it can be seen as the expression of a curve, then
Referred to as one linear connection subset.
Apparatus above embodiment and embodiment of the method are one-to-one, the simple places of Installation practice, referring to method reality
Apply example.
Each embodiment in this specification is described in a progressive manner, the highlights of each of the examples are with other
The difference of embodiment, the same or similar parts in each embodiment may refer to each other.
Professional further appreciates that, unit described in conjunction with the examples disclosed in the embodiments of the present disclosure
And algorithm steps, can be realized with electronic hardware, computer software, or a combination of the two, in order to clearly demonstrate hardware and
The interchangeability of software generally describes each exemplary composition and step according to functionality in the above description.This
A little functions are implemented in hardware or software actually, the specific application and design constraint depending on technical solution.Specially
Industry technical staff can use different methods to achieve the described function each specific application, but this realization is not
It should be more than the scope of the present invention.
The step of method described in conjunction with the examples disclosed in this document or algorithm, can directly be held with hardware, processor
The combination of capable software module or the two is implemented.Software module can be placed in random access memory, memory, read-only memory,
Electrically programmable ROM, electricity can sassafras except in programming ROM, register, hard disk, moveable magnetic disc, CD-ROM or technical field institute it is public
In the storage medium for any other forms known.
The embodiment of the present invention is described with above attached drawing, but the invention is not limited to above-mentioned specific
Embodiment, the above mentioned embodiment is only schematical, rather than restrictive, those skilled in the art
Under the inspiration of the present invention, without breaking away from the scope protected by the purposes and claims of the present invention, it can also make very much
Form, all of these belong to the protection of the present invention.